Thomas/re enable tests#63
Merged
Merged
Conversation
…nable-tests # Conflicts: # blast/manipulator/Kinova_Gen3.hpp # blast/manipulator/UR5e.hpp # examples/CMakeLists.txt # tests/CMakeLists.txt
- moved to utililites - moved scenes.hpp to blast/world/
…t into Thomas/Re-enable-tests
…t into Thomas/Re-enable-tests
random_int used a shared static std::mt19937, which is a data race when guesses are generated in parallel via Taskflow (mt19937::operator() mutates state). Switch rd/gen to static thread_local to match get_random() above; each thread seeds its own generator once. The distribution stays local so the [min, max] range can vary per call.
Clearer naming alongside random_int. Updates the forward declaration in blast_math.hpp, the definition in math/misc.hpp, and all call sites in math/Matrix.hpp, math/Array.hpp, and the gpu/ trajectory and manipulator headers. get_random_guesses is a separate function and is unchanged.
The dev preset hardcoded CMAKE_CXX_FLAGS=-march=native and relied on a condition to gate per OS. That conflates OS with compiler: on Windows the flag depends on the toolchain (cl/clang-cl want /arch:AVX2; MinGW GCC and Clang's GNU driver want -march=native), and a preset condition cannot inspect the detected compiler. The arch flag is already applied correctly per-target as PRIVATE in examples/ and tests/ (/arch:AVX2 for MSVC, -march=native otherwise), so the global preset flag was a redundant, broken duplicate. Drop it and the OS conditions, and merge dev-msvc into a single dev preset that is correct for MSVC, GCC, Clang and MinGW on any OS. Add configuration:RelWithDebInfo to the dev build preset so multi-config generators (Visual Studio) build the intended config instead of the default.
Blast's headers call AVX2/FMA intrinsics directly (e.g. _mm256_fmadd_pd in math/Array.hpp), but the arch flags were never propagated to consumers and there was no guard, so find_package(Blast) users hit a cryptic 'target specific option mismatch' on GCC/Clang (or silent degradation on MSVC). The CI smoke test only built a Vec3, so it never caught this. - blast/blast: add an #error guard that reports a clear, per-compiler message when AVX2 (and FMA, off MSVC) is missing. CUDA builds opt out. - blast/CMakeLists.txt: propagate -mavx2 -mfma as INTERFACE for GCC/Clang. Feature flags are additive, so they never downgrade a consumer's -march=native. MSVC is excluded (its /arch: is mutually exclusive and would downgrade an AVX512 consumer with D9025); it relies on the guard. - CI: the consumer smoke test now calls blast::dot, exercising the AVX2/FMA path with no flags set, so it verifies the propagated contract. - CLAUDE.md: document the requirement and correct the stale claim that blast/CMakeLists.txt sets these flags.
Resolves the modify/delete conflict on tests/benchmarks/bench_functions.cpp by keeping the deletion: the file references test_helper/ which was removed as part of this branch.
gallantandre
approved these changes
Jun 14, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.